2014 VGDRA

526 IEEE SENSORS JOURNAL, VOL. 15, NO. 1, JANUARY 2015

VGDRA: A Virtual Grid-Based Dynamic RoutesAdjustment Scheme for Mobile Sink-Based

Wireless Sensor NetworksAbdul Waheed Khan, Abdul Hanan Abdullah, Member, IEEE,

Mohammad Abdur Razzaque, Member, IEEE, and Javed Iqbal Bangash

Abstract In wireless sensor networks, exploiting the sinkmobility has been considered as a good strategy to balancethe nodes energy dissipation. Despite its numerous advantages,the data dissemination to the mobile sink is a challenging taskfor the resource constrained sensor nodes due to the dynamicnetwork topology caused by the sink mobility. For efficientdata delivery, nodes need to reconstruct their routes towardthe latest location of the mobile sink, which undermines theenergy conservation goal. In this paper, we present a virtual grid-based dynamic routes adjustment (VGDRA) scheme that aimsto minimize the routes reconstruction cost of the sensor nodeswhile maintaining nearly optimal routes to the latest location ofthe mobile sink. We propose a set of communication rules thatgoverns the routes reconstruction process thereby requiring onlya limited number of nodes to readjust their data delivery routestoward the mobile sink. Simulation results demonstrate reducedroutes reconstruction cost and improved network lifetime of theVGDRA scheme when compared with existing work.

Index Terms Routes reconstruction, energy efficiency, mobilesink, wireless sensor networks.

I. INTRODUCTION

W IRELESS Sensor Network (WSN) a self-organizednetwork of tiny computing and communication devices(nodes) has been widely used in several un-attended anddangerous environments. In a typical deployment of WSN,nodes are battery operated where they cooperatively mon-itor and report some phenomenon of interest to a centralnode called sink or base-station for further processing andanalysis. Traditional static nodes deployment where nodesexhibit n-to-1 communication in reporting their observed datato a single static sink, gives rise to energy-hole phenomenonin the vicinity of sink. Sink mobility introduced in [1] notonly helps to balance the nodes energy dissipation but canalso link isolated network segments in problematic areas [2].

Manuscript received June 2, 2014; revised July 27, 2014; acceptedJuly 27, 2014. Date of publication August 12, 2014; date of current versionNovember 11, 2014. This work was supported by the Universiti TeknologiMalaysia, Malaysia, under Grant PY/2012/00306 (Vote No: 4F205). Theassociate editor coordinating the review of this paper and approving it forpublication was Dr. Nitaigour P. Mahalik.

The authors are with the Department of Computer Science, Faculty ofComputing, Universiti Teknologi Malaysia, Johor Bahru 81310, Malaysia(e-mail: [email protected]; [email protected]; [email protected];[email protected]).

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/JSEN.2014.2347137

In addition, several application environments naturally requiresink mobility in the sensor field [3] e.g., in a disaster man-agement system, a rescuer equipped with a PDA can movearound the disaster area to look for any survivor. Similarly,in a battlefield environment, a commander can obtain real-time information about any intrusion of enemies, scale ofattack, suspicious activities etc via field sensors while on themove. In an Intelligent Transport System (ITS), sensor nodesdeployed at various points of interest - junctions, car parks,areas susceptible to falling rocks, can provide early warningsto drivers (mobile sink) well ahead of their physical approach.

Exploiting the sinks mobility helps to prolong the networklifetime thereby alleviating energy-hole problem; however, itbrings new challenges for the data dissemination process.Unlike static sink scenarios, the network topology becomesdynamic as the sink keeps on changing its location [4].To cope with the dynamic network topology, nodes needto keep track of the latest location of the mobile sink forefficient data delivery. Some data dissemination protocolse.g., Directed Diffusion [5], propose periodic flooding of sinkstopological updates in the entire sensor field which gives riseto more collisions and thus more retransmissions. Taking intoconsideration the scarce energy resource of nodes, frequentpropagation of sinks mobility updates should be avoided asit greatly undermines the energy conservation goal. In thisregard, to enable sensor nodes to maintain fresh routes towardsthe mobile sink while incurring minimal communication cost,overlaying based virtual infrastructure over the physical net-work is considered as an efficient approach [6]. In the virtualinfrastructure based data dissemination schemes, only a set ofdesignated nodes scattered in the sensor field are responsibleto keep track of sinks location. Such designated nodes gatherthe observed data from the nodes in their vicinity during theabsence of the sink and then proactively or reactively reportdata to the mobile sink.

In this paper, a novel scheme called Virtual Grid basedDynamic Routes Adjustment (VGDRA) is proposed for peri-odic data collection from WSN. Unlike the existing solutions,which improve data delivery performance either by employingmultiple mobile sinks or by deploying super nodes at strategi-cally important points in the sensor field, the proposed schemedoes not impose any such constraints. It aims to optimize thetrade-off between nodes energy consumption and data delivery

1530-437X 2014 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted,but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

KHAN et al.: VGDRA SCHEME FOR MOBILE SINK-BASED WSNs 527

performance using a single mobile sink while adhering tothe low-cost theme of WSN. The proposed scheme enablessensor nodes to maintain nearly optimal routes to the latestlocation of a mobile sink with minimal network overhead.It partitions the sensor field into a virtual grid of K equal sizedcells and constructs a virtual backbone network comprised ofall the cell-headers. Nodes close to the centre of the cellsare appointed as cell-headers, which are responsible for datacollection from member nodes within the cell and deliveringthe data to the mobile sink using the virtual backbone network.The goal behind such virtual structure construction is to mini-mize the routes re-adjustment cost due to sink mobility so thatthe observed data is delivered to the mobile sink in an energyefficient way. In addition, VGDRA also sets up communicationroutes such that the end-to-end delay and energy cost isminimized in the data delivery phase to the mobile sink.The mobile sink moves along the periphery of the sensorfield and communicates with the border cell-headers for datacollection. The routes re-adjustment process is governed by aset of rules to dynamically cope with the sink mobility. UsingVGDRA, only a subset of the cell-headers needs to take part inre-adjusting their routes to the latest location of the mobile sinkthereby reducing the communication cost. Simulation resultsreveal decreased energy consumption and faster convergenceof VGDRA compared to other state-of-the art.

The rest of this paper is organized as follows: Section IIdescribes the related work encompassing the variousapproaches in literature that deal with data delivery to a mobilesink in WSN. Section III presents our VGDRA scheme indetail. To evaluate the performance of the VGDRA scheme,simulation setup and results are presented in Section IV.Finally, Section V concludes the paper.

II. RELATED WORK

Several virtual infrastructure based data dissemination pro-tocols have been proposed for mobile sink based WSN inthe last decade. Based on the mobility pattern exhibited bythe sink in the sensor field, the data collection or dissemi-nation schemes can be classified into controlled and uncon-trolled sink mobility schemes. In controlled sink mobilityschemes [7]-[10], the mobility (speed and/or direction) ofthe sink is manipulated and controlled either by an externalobserver or in accordance with the network dynamics. Theuncontrolled sink mobility based schemes are characterizedby the fact that the sink makes its next move autonomouslyin terms of speed and direction. This paper considers theuncontrolled sink mobility environments and in the followinglines, we briefly describe the related works in this contextincluding their methodology and the relative strengths andweaknesses.

Chen et al. [11] presented a converge-cast tree algorithmcalled Virtual Circle Combined Straight Routing (VCCSR)that constructs a virtual structure comprised of virtual circlesand straight lines. A set of nodes are appointed as cluster-heads along these virtual circles and straight lines. Togetherthe set of cluster-heads form a virtual backbone network. Thesink circulates the sensor field and maintains communication

with the border cluster-heads for data collection. The cluster-heads in VCCSR follow a set of communication rules tominimize the routes re-adjustment cost in propagating thesinks latest location information. VCCSR scheme althoughreduces the routes reconstruction cost in handling the sinkmobility, however, the cluster-head at the centre of the sensorfield being the focal point in routes re-adjustment process,depletes its energy much earlier.

Hexagonal cell-based Data Dissemination (HexDD) pro-posed in [12] constructs a hexagonal grid structure to addressreal-time data delivery while taking into consideration thedynamic conditions of multiple mobile sinks and eventsources. Based on the six directions of a hexagon, HexDDdefines query and data rendezvous lines to avoid redundantpropagation of sinks data queries. Nodes send their data tonearest border line which is then propagated towards the centrecell. Nodes along the border line store and replicate the data.Sinks data queries are forwarded towards the centre cell andas soon as it approaches a border line node with the relevantdata stored, data delivery to the mobile sink starts using thereverse path. To cope with sink mobility, whenever the sinkmoves from one cell to another, it informs the centre nodesas well as the border nodes along the route about the newcell where the sink is currently stationed. This results in highenergy consumption especially at higher sinks speeds. Nodesalong the border line cells and especially at the centre cell arevulnerable to high energy consumption thereby causing earlyhot-spot problem.

Oh et al. proposed a data dissemination scheme calledBackbone-based Virtual Infrastructure (BVI) in [13] thatmakes use of single-level multi-hop clustering. It aims tominimize the total number of clusters and thus the scale ofnetwork overhead associated with informing all the CH nodesabout the sinks location information. For clustering it employsHEED [14] where priority is given to residual energy level ofnodes in electing the CH nodes. To keep track of sink locationinformation, it assumes that the network operator appoints acertain CH node as root of the tree. Whenever, the mobilesink joins the sensor field, it registers itself with the closestCH via an agent node. The host CH node accordingly updatesother CH nodes along the route to the root CH about thesinks location information. Furthermore, when a mobile sinkmoves within a cluster, the respective CH node only takescare of connection with the sink within the cluster and avoidspropagation of sink location updates to the root. However,when the sink joins another cluster, it selects another agentnode and registers itself to the new CH node which accordinglyshares this information with the root and the other CH nodesalong the BVI segment to the root. The multi-hop clusteringalthough appears a good strategy to minimize the number ofclusters and thus the network control overhead, however, theroot node being the focal point in routes adjustments triggersearly energy depletion and thus reduces the network lifetime.

Multiple Enhanced Specified-deployed Sub-sinks (MESS)in [15], creates a virtual strip in the middle of sensor fieldthereby placing enhanced wireless nodes (sub-sinks) havingmore storage capacity at equal distances. The set of sub-sinknodes along the accessible path serve as rendezvous points for


TABLE ISUMMARY AND COMPARISON OF VIRTUAL STRUCTURE BASED DATA DISSEMINATION PROTOCOLS

the mobile sink and collect and store data from sensor nodes.In data delivery phase, mobile sink floods the query alongthe virtual strip till it reaches to the sub-sink node owning thedata. Upon receiving the query from mobile sink, the sub-sinksroute their deposited data to the mobile sink using geograph-ical forwarding approach. A similar approach has also beenproposed in Line-Based Data Dissemination (LBDD) [16]which constructs a vertical line by dividing the sensor fieldinto two equal sized blocks. Yet another similar approachcan be found in [17], which places a virtual rail (calledRailRoad) in the middle of the sensor field where nodesinside the virtual rails premises serve as rendezvous points.The main limitation of MESS, LBDD, and RailRoad is theearly energy depletion of nodes close to the virtual structureas the same nodes are repeatedly chosen as relays for thefarther nodes. In addition, MESS also imposes placementof enhanced nodes along the virtual strip which limits itsapplicability.

In Quadtree-based Data Dissemination (QDD) proposed byMir and Ko in [18], a node upon detecting an event calculates aset of rendezvous points (RPs) by successively partitioning thephysical network space into four quadrants of uniform sizes.After partitioning the network, QDD routes the observed datato those nodes which are close to the centroid of each partition.The mobile sink disseminates the query packet using the samestrategy by querying the node at closest RP first, followedby the subsequent RP nodes till it reaches the required datareport. In static nodes deployments, the same set of nodesbecome RPs repeatedly which results in early energy deple-tion of those nodes and thus decreases the overall networklifetime.

Virtual grid based Two-Tier Data Dissemination (TTDD)in [19] proactively constructs a uniform per source nodevirtual grid structure spanning the entire sensor field. For datacollection, the mobile sink floods its local grid cell wherethe query packet makes use of all the disseminating pointsalong the virtual grid till it gets to the source node. Duringquery dissemination process, a reverse path is also establishedfor data reporting to the mobile sink. TTDD although avoidsthe flooding of the sinks topological updates, however, theper source virtual grid construction undermines the networklifetime.

Geographical Cellular-like Architecture (GCA) in [20]proactively constructs a cellular-like hierarchical hexagonal

virtual structure for handling sink mobility. GCA just likehome-agent in cellular networks appoints the node close tothe centre of the cell as the header node and remark therest of the nodes as member nodes. For data collection, themobile sink sends its query to nearest header which thenpropagates the query to all the headers. To handle sinkmobility, when the sink joins another cell, it informs theold cells header about the new header which re-route thepackets accordingly. GCA although avoids flooding of sinkslocation information, however, the non-optimal data deliverypaths results in increased latency and packet loss ratio due tothe expiry of time-to-live value of packets.

Hierarchical Cluster-based Data Dissemination (HCDD)in [21] proposes a hierarchical cluster architecture where thesecond level cluster-heads of the mobile sink are appointedas routing agents. The routing agents are responsible to keeptrack of sinks latest location information and all the cluster-heads route their collected data to the nearby routing agents.When sink moves from one point to another, it informsthe nearest routing agent via the closest cluster-head. Therouting agent upon sink discovery broadcasts the sinks latestlocation information to all the other routing agents. In highsink mobility, nodes using HCDD suffer from high energyconsumption. In addition, due to the restricted propagation ofsinks location information, the data delivery paths are notoptimal which results in high latency.

Table 1 presents a summary and comparison of the vari-ously discussed schemes. We evaluate each scheme in termsof parameters such as location awareness, cost of networkcontrol overhead, convergence time and applicability. Thelocation awareness specifies whether the considered schemerequires the nodes to be aware of their physical/relativecoordinates. This feature is quite helpful in constructing thevirtual infrastructure as well as routing of the query andresponse packets [22], [23]; however it incurs some additionalenergy cost. The second parameter provides an estimate ofthe overhead cost involved in establishing and maintainingdata delivery routes to the latest location of the mobile sink.For prompt delivery of data packets to the mobile sink, nodesneed to be informed of the latest location of the mobile sink.Similarly, the convergence time is an estimate of how promptlythe sensor nodes come to know about some significant positionchange of the mobile sink and indirectly reflects the efficiencyof the subsequent data delivery phase. Finally, the applicability


parameter represents the scope of each scheme where ifthe considered scheme imposes too many constraints and/orviolates the self-organized nature of the low cost sensor nodes,it limits its widespread applicability.

Based on the study of literature, it is observed that improveddata delivery performance can be achieved at the expense ofmore energy consumption in the form of frequent propagationof sinks location updates. The main contribution of thispaper is to optimize the trade-off between nodes energyconsumption and improved data delivery performance therebyproposing a virtual grid based dynamic routes adjustmentscheme. It offers a light weight solution for the resourceconstrained real motes which neither requires any specializedset of resources on part of the sensor nodes nor imposestoo many constraints for network operation. These featurestogether with the dynamic routes adjustments in an energy-efficient manner make it a viable choice in a range of WSNsapplications as compared to existing schemes.

III. THE VGDRA SCHEMEIn this section, we give detailed description of our VGDRA

scheme, including how to construct the virtual infrastructureand how to maintain fresh routes towards the latest locationof the mobile sink. We design a virtual infrastructure bypartitioning the sensor field into a virtual grid of uniformsized cells where the total number of cells is a functionof the number of sensor nodes. A set of nodes close tocentre of the cells are appointed as cell-headers which areresponsible for keeping track of the latest location of themobile sink and relieve the rest of member nodes from takingpart in routes re-adjustment. Nodes other than the cell-headersassociate themselves with the closest cell-headers and reportthe observed data to their cell-headers. Adjacent cell-headerscommunicate with each other via gateway nodes. The set ofcell-headers nodes together with the gateway nodes constructsthe virtual backbone structure.

A. Network CharacteristicsBefore describing the methodology of VGDRA scheme, it is

worthwhile to highlight the various assumptions of the sensornetworks. We assume the following network characteristics:

Nodes are randomly deployed and throughout remainstatic.

All the nodes are of homogeneous architecture and knowtheir location information.

Nodes adapt their transmission power based on the dis-tance to the destination nodes.

The mobile sink does not have any resources constraints. The mobile sink performs periodic data collection from

sensor nodes while moving along the periphery of thesensor field and maintains communication with the closestborder-line cell-headers for data collection.

B. The Virtual Structure ConstructionThe VGDRA scheme constructs the virtual grid structure

by first partitioning the sensor field into several uniform sized

Fig. 1. Example of different virtual grid based structures for different numberof nodes

cells based on the number of nodes in the sensor field. Therationale behind such portioning is to uniformly distribute thework-load on part of cell-header nodes which consequentlyresults in prolonged network lifetime.

To determine the optimal number of cells and thus thecluster-heads, we adopt the heuristics used in LEACH [24],TEEN [25], and APTEEN [26] which consider 5% of thetotal number of sensor nodes. Given N number of nodes, theVGDRA scheme partitions the sensor field into K uniformsized cells using Equation 1, where K is a squared number.Fig. 1 (a), (b) and (c) shows network partitioning into variousuniform sized cells for N = 100, 200, 300 respectively.

K =

4916...

N 0.05 66 < N 0.05 1212 < N 0.05 20...

(1)

After the network partitioning, next VGDRA schemeappoints a set of nodes as cell-headers. Initially in every cell,the node closest to the mid-point of the cell is elected asthe cell-header. Nodes using the knowledge of sensor fieldsdimension and the total number of nodes compute the mid-points of all the cells. In order to reduce the communicationcost in the cell-header election, only those nodes take partin the election whose distance to the mid-point of the cell isless than a certain threshold. The threshold distance to themid-point is gradually increased if no node can be foundwithin the threshold distance around the mid-point of the cell.This threshold based cell-header election strategy not onlyhelps in energy conservation but also elects the cell-headerat the most appropriate position within the cell. After theinitial cell-header election, each cell-header notifies its statusnot only to the surrounding nodes within its cell but alsoto the nodes which are slightly beyond the cell boundary.Nodes might receive cell-header notifications from more thanone cell-header and associate themselves to the closest one.Nodes that receive notifications from multiple cell-headers alsoshare the information of the secondary cell-header with theirprimary cell-header. In this way, each cell-header form adja-cencies with neighboring cell-headers using gateway nodes.The maximum number of adjacent cell-headers for a border-line cell-header is 3 whereas for an inside cell-header is 4.The set of cell-header nodes together with the gateway nodesconstructs a chain like virtual backbone structure as shownin Fig. 2.

After the cell-header election and establishing the adjacen-cies, communication routes are setup considering the mobile


Fig. 2. An example of virtual backbone structure after establishingadjacencies.

Fig. 3. An example of virtual backbone structure after initial routes setup.

sink is located at coordinates (0, 0). As a result of the initialroutes setup, all the cell-headers adjust their routes to the initialposition of the mobile sink. Fig. 3 shows the virtual backbonestructure after the initial routes setup when the sensor field ispartitioned into 16 cells.

C. Dynamic Routes AdjustmentIn order to cope with dynamic network topology caused by

sink mobility, nodes need to setup their data delivery routesin accordance with the latest location of the mobile sink.Flooding the sinks latest location to the entire sensor fieldis the most naive approach in this regard but greatly under-mines the energy conservation goal and is therefore avoided.Using our VGDRA scheme, only the set of cell-headers thatconstitute the virtual backbone structure are responsible formaintaining fresh routes to the latest location of mobile sink.For periodic data collection from the sensor field, the mobilesink moves around the sensor field and collects data via theclosest border-line cell-header. The closest border-line cell-header (originating cell-header) upon discovering the sinkspresence, shares this information with the rest of the cell-headers in a controlled manner. The VGDRA scheme definesa set of propagation rules so that only those cell-headers takepart in the routes re-adjustment process that really require

Fig. 4. An example of routes re-adjustments when sink moves from cell 2to cell 3.

to adjust their routes. The propagation rules are described asfollows:

Rule 1: The originating cell-header upon sink discoveryfirst verifies whether its next-hop is already set to the mobilesink or not. If the mobile sink was previously being setup asits next-hop, the originating cell-header does not propagatesinks location update. However, if the next-hop entry ofthe originating cell-header is other than the mobile sink, itexercises rule 2.

Rule 2: The originating cell-header being one-hop from themobile sink sets the mobile sink as its next-hop and shares thisinformation with the previous originating cell-header and itsdownstream adjacent cell-header.

Rule 3: The previous originating cell-header upon receivingthe sinks location update from the current originating cell-header, adjusts its data delivery route by setting the currentoriginating cell-header as its next-hop towards the sink.

Rule 4: The downstream cell-header upon receiving thesinks location update checks whether the sender cell-headeris the same as its previous next-hop or different. If it is thesame, the downstream cell-header drops the sinks locationupdate packet and does not propagate it further to the nextdownstream cell-header. In the case when it is different, thedownstream cell-header updates its next-hop entry to the newsender cell-header and further propagates the sinks locationupdate to the next downstream cell-header. This procedure isrepeated till all the downstream cell-headers adjust their datadelivery routes towards the latest location of the mobile sink.

Fig. 4(a) shows an example of the data delivery paths whenthe sink is located in the cell 2 premises. When the mobile sinkmoves from cell 2 to cell 3, the cell-header at cell 3 exercisesrule 2 and rule 3 to update the cell-header at cell 2, followed byrule 4 to update its downstream cell-headers i.e., 7, 11 and 15as shown in Fig. 4(b). In this way, only a limited numberof cell-headers take part in the routes re-adjustment processthereby reducing the overall routes re-adjustment cost of thenetwork.

Similarly, Fig. 5 demonstrates when the mobile sink movesfrom position a to b within the same cell, the cell-headerat cell 4 exercises rule 1 and refrains itself from propagatingsinks location information. This strategy helps to minimize the


Algorithm 1 Routes Re-Adjustment Using VGDRA Scheme1. Mobile Sink (MS) updates its location to the closest

Cell-Header (CH).2. The closest CH becomes Originating Cell-Header

(OCH).3. if the previous Next_Hop of OCH is not the MS4. {5. set Next_Hop of OCH MS6. OCH sends route update packet to the previous OCH7. set Next_Hop of previous OCH OCH8. OCH sends route update packet to its immediate

downstream CH9. for each downstream CH receives route update packet

10. {11. if the previous Next_Hop of CH is not the current

sender12. {13. set Next_Hop of CH current sender14. if next downstream CH is not NULL15. {16. set sender current CH17. Current CH sends route update packet to its

immediate downstream CH18. }19. else20. drop the packet21. }22. else23. drop the packet24. }25. }26. else27. drop the packet

Fig. 5. An example of preventing the undesired propagation of sink locationupdates.

routes reconstruction cost to a great extent and thus improvesthe network lifetime.

The VGDRA algorithm that governs the routes adjustmentprocess along the sink mobility is described in detail as shownin top of the page.

D. Cell-Header RotationAn integral part of the proposed VGDRA scheme is rotating

the role of the cell-header in every cell. The cell-header beingthe local data collector is vulnerable to high energy dissipationand therefore to prolong the network lifetime, the cell-headerrole needs to be distributed among the nodes within the cell.In order to achieve uniform energy dissipation, the VGDRAscheme keeps track of the residual energy level of the currentcell-header, where if it gets below a certain threshold, thenew cell-header election is initiated by the current cell-header.In the re-election process, the node that is relatively moreclose to the mid-point of the cell and has a higher energylevel compared to other candidates is elected as the newcell-header. Also in the re-election process, the search zonearound the mid-point in every cell is slightly increased or theenergy threshold level is decreased progressively if no suitablenode can be found. In order to preserve the virtual backbonestructure, the current cell-header before stepping down, sharesthe information of the new cell-header not only with all itsmember nodes but also with the adjacent cell-headers in itsneighborhood.

IV. SIMULATION AND RESULTSIn this section, we present the simulation results using NS-2.

We varied the total number of sensor nodes from 100 to 400which are randomly deployed in a sensor field of 200 200dimension. A mobile sink moves around the sensor fieldcounterclockwise and periodically broadcasts hello packets.Initially all the sensor nodes have uniform energy reserveof 1 mJ. We considered the energy model being used in [27]and assumed free space radio propagation model (d2, d isthe distance between sender and receiver). Furthermore, weconsidered nodes energy consumption in transmission (Tx)and receiving (Rx) modes only which are computed usingEquation 2 and 3 respectively.

Tx = (Eelect K ) +(EampK d2

) (2)Rx = Eelect K (3)

In Equation 2 and 3, K is the message length, Eelect is thenodes energy dissipation in order to run its radio electroniccircuitry and Eamp is the energy dissipation by the transmitteramplifier to suppress the channel noise. In our experiment, wetook Eelect = 50 nJ, and Eamp = 10 nJ/bit/m2 and K = 8 bits.We considered the nodes communication cost in adjusting thedata delivery routes only, whereas the actual data delivery isbeyond the scope of this paper.

We compared our VGDRA scheme with VCCSR, HexDD,and BVI where a common feature among them is the use ofa virtual infrastructure for network operation. We used fourdifferent criteria to evaluate the performance of the VGDRAagainst the other schemes under the same network dynamics:virtual backbone structure construction cost, per round routesreconstruction cost, average network lifetime, and networkconvergence time.

A. The Virtual Backbone Structure Construction CostThe virtual structure construction cost is an estimate of

the nodes energy consumption in electing the cell-headers


Fig. 6. Comparing the virtual structure construction cost for different networksizes.

and then forming the virtual backbone network. Fig. 6 com-pares the average nodes energy consumption of our VGDRAscheme with the other schemes in constructing the virtualbackbone network for different network sizes.

As demonstrated in Fig. 6, nodes using VGDRA schemeincur least cost compared to other schemes in constructingthe virtual structure. The VCCSR considers fixed number ofcluster-head nodes irrespective of the network size e.g., itconsiders 81 cluster-head nodes under the considered networkdynamics and thus as a result, a high population of the sensornodes take part in the cluster-head election. Similarly, the BVIincurs considerable communication cost in clustering the net-work where all the nodes exchange residual energy level infor-mation. Compared to VCCSR and BVI, nodes using HexDDperform local processing thereby causing less communicationoverhead. On contrary, using our VGDRA scheme, the totalnumber of cells and thus the cell-headers is a function of thetotal number of nodes e.g., the number of cell-headers variesfrom 4 to 16 when N varies from 100 to 400 nodes. In addi-tion, only the nodes within short distance to the mid-point ofthe cell take part in cell-header election thereby reducing thecommunication cost.

B. The Per Round Routes Reconstruction CostThe per round routes reconstruction cost represents the

nodes energy expenditure in re-adjusting the data deliveryroutes as the sink moves around the sensor field and completesone round of the sensor field. As shown in Fig. 7, usingthe VGDRA scheme, the average nodes energy consump-tion in reconstructing the data delivery routes to the latestlocation of the mobile sink is significantly less comparedto the other schemes. This is mainly attributed to the leastpropagation of sinks location updates by following the setof communication rules of the VGDRA while preservingnearly optimal routes towards the latest location of the mobilesink. Using our VGDRA scheme, only a partial sub-setof cell-header nodes takes part in the routes reconstructionprocess thereby reducing the overall routes reconstructioncost as the mobile sink completes one round of the sensorfield.

Fig. 7. Comparing the per round routes reconstruction cost for differentnetwork sizes.

Fig. 8. Comparing the network lifetime in terms of number of rounds aroundthe sensor field.

C. The Network LifetimeThe network lifetime is defined as the time elapsed since the

nodes deployment till the first node dies due to energy deple-tion. In our experiments, we estimated the network lifetimein terms of the number of rounds of the mobile sink aroundthe sensor field till the first node in the network dies due toenergy depletion. As presented in Fig. 8, our VGDRA schemeoutperforms the other schemes in terms of network lifetime atdifferent network sizes. In VCCSR, the cluster-head at thecentral-point of the sensor field suffers from more work-loadfor taking part in every single reconstruction phase and thusdepletes its energy much earlier compared to others. Similarbehavior is exhibited by the centre and border nodes in HexDDthereby decreasing the overall network lifetime. Unlike theVCCSR and HexDD, the proposed VGDRA and BVI schemeskeep track of the residual energy of cell-header nodes andprogressively elect new header nodes thereby prolonging thenetwork lifetime. Furthermore, compared to BVI, the proposedVGDRA scheme incurs least network control overhead. Theresults presented in Fig. 8 also demonstrates nearly uniformnetwork lifetime at different network sizes using our VGDRAscheme which justifies our approach of partitioning the sensorfield into different number of cells on the basis of the totalnumber of nodes.


Fig. 9. Comparing the network convergence time for different network sizes.

D. The Network Convergence TimeThe network convergence time is an indirect reflection of

the data delivery efficiency as the more promptly the nodescome to know about the latest location of a mobile sink,the more efficient routes they can select in disseminating thesensed data. It is an estimation of the elapsed time that asignificant position change of the mobile sink is recordedby the nodes constituting the virtual infrastructure. In termsof convergence time, the faster the nodes converge to latestlocation of a mobile sink, the better they perform in datadissemination phase. As shown in Fig. 9, the convergencetime of the proposed VGDRA scheme is very fast comparedto VCCSR, HexDD, and BVI when the sink is moving ata speed of 10 m/sec. Using the set of communication rules,our VGDRA scheme intelligently picks a small subset of cell-headers in the routes re-adjustment process and then greedilyshares the latest location information of the mobile sink withthem. This partial re-adjustment greatly reduces the networkoverhead and leads to faster convergence of nodes to the latestlocation of the mobile sink.

V. CONCLUSION AND FUTURE WORKIn this paper, we proposed a novel Virtual Grid based

Dynamic Routes Adjustment (VGDRA) scheme that incursleast communication cost while maintaining nearly optimalroutes to the latest location of the mobile sink. Our VGDRAscheme partitions the sensor field into a virtual grid andconstructs a virtual backbone structure comprised of the cell-header nodes. A mobile sink while moving around the sensorfield keeps on changing its location and interacts with theclosest border-line cell-header for data collection. Using a setof communication rules, only a limited number of the cell-headers take part in the routes reconstruction process therebyreducing the overall communication cost. In terms of nodesenergy consumption, the simulation results reveal improvedperformance of our VGDRA scheme for different networksizes.

Considering the scope of this paper, we have not includedthe actual data delivery model. In future work, we will analyzethe performance of our VGDRA scheme at different sinks

speeds and different data generation rates of the sensor nodes.The proposed VGDRA scheme though offers a light weightsolution and does not impose many constraints on part of theresource constrained sensor motes, yet its practical implemen-tation on real hardware needs to be confirmed. We also aimto develop a small test bed for the practical implementation ofthe proposed VGDRA scheme on real hardware (motes) andevaluate its results.

REFERENCES

[1] R. C. Shah, S. Roy, S. Jain, and W. Brunette, Data MULEs: Modelingand analysis of a three-tier architecture for sparse sensor networks, inAd Hoc Netw., vol. 1. 2003, pp. 215233.

[2] S. R. Gandham, M. Dawande, R. Prakash, and S. Venkatesan, Energyefficient schemes for wireless sensor networks with multiple mobilebase stations, in Proc. IEEE Global Telecommun. Conf. (GLOBECOM),vol. 1. Dec. 2003, pp. 377381.

[3] A. W. Khan, A. H. Abdullah, M. H. Anisi, and J. I. Bangash, A compre-hensive study of data collection schemes using mobile sinks in wirelesssensor networks, Sensors, vol. 14, no. 2, pp. 25102548, 2014.

[4] M. Di Francesco, S. K. Das, and G. Anastasi, Data collection inwireless sensor networks with mobile elements, ACM Trans. SensorNetw., vol. 8, no. 1, pp. 131, Aug. 2011.

[5] I. Chalermek, R. Govindan, and D. Estrin, Directed diffusion: A scal-able and robust communication paradigm for sensor networks, in Proc.ACM SIGMOBILE Int. Conf. Mobile Comput. Netw. (MOBICOM), 2000,pp. 5667.

[6] E. B. Hamida and G. Chelius, Strategies for data dissemination tomobile sinks in wireless sensor networks, IEEE Wireless Commun.,vol. 15, no. 6, pp. 3137, Dec. 2008.

[7] A. Kinalis, S. Nikoletseas, D. Patroumpa, and J. Rolim, Biased sinkmobility with adaptive stop times for low latency data collection insensor networks, Inf. Fusion, vol. 15, pp. 5663, Jan. 2014.

[8] W. M. Aioffi, C. A. Valle, G. R. Mateus, and A. S. da Cunha, Balancingmessage delivery latency and network lifetime through an integratedmodel for clustering and routing in wireless sensor networks, Comput.Netw., vol. 55, no. 13, pp. 28032820, Sep. 2011.

[9] B. Nazir and H. Hasbullah, Mobile sink based routing protocol (MSRP)for prolonging network lifetime in clustered wireless sensor network,in Proc. Int. Conf. Comput. Appl. Ind. Electron. (ICCAIE), Dec. 2010,pp. 624629.

[10] T. Banerjee, B. Xie, J. H. Jun, and D. P. Agrawal, Increasing lifetimeof wireless sensor networks using controllable mobile cluster heads,Wireless Commun. Mobile Comput., vol. 10, no. 3, pp. 313336,Mar. 2010.

[11] T.-S. Chen, H.-W. Tsai, Y.-H. Chang, and T.-C. Chen, Geographicconvergecast using mobile sink in wireless sensor networks, Comput.Commun., vol. 36, no. 4, pp. 445458, Feb. 2013.

[12] A. Erman, A. Dilo, and P. Havinga, A virtual infrastructure based onhoneycomb tessellation for data dissemination in multi-sink mobile wire-less sensor networks, EURASIP J. Wireless Commun. Netw., vol. 2012,no. 17, pp. 154, 2012.

[13] S. Oh, E. Lee, S. Park, J. Jung, and S.-H. Kim, Communicationscheme to support sink mobility in multi-hop clustered wireless sensornetworks, in Proc. 24th IEEE Int. Conf. Adv. Inf. Netw. Appl., Apr. 2010,pp. 866872.

[14] O. Younis and S. Fahmy, HEED: A hybrid, energy-efficient, distributedclustering approach for ad hoc sensor networks, IEEE Trans. MobileComput., vol. 3, no. 4, pp. 366379, Oct. 2004.

[15] B. Tang, J. Wang, X. Geng, Y. Zheng, and J.-U. Kim, A novel dataretrieving mechanism in wireless sensor networks with path-limitedmobile sink, nt. J. Grid Distrib. Comput., vol. 5, no. 3, pp. 133140,2012.

[16] E. B. Hamida and G. Chelius, A line-based data dissemination protocolfor wireless sensor networks with mobile sink, in Proc. IEEE Int. Conf.Commun., May 2008, pp. 22012205.

[17] J.-H. Shin and D. Park, A virtual infrastructure for large-scalewireless sensor networks, Comput. Commun., vol. 30, nos. 1415,pp. 28532866, Oct. 2007.

[18] Z. H. Mir and Y.-B. Ko, A quadtree-based data dissemination protocolfor wireless sensor networks with mobile sinks, in Proc. PersonalWireless Commun., 2006, pp. 447458.


[19] H. Luo, F. Ye, J. Cheng, S. Lu, and L. Zhang, TTDD: Two-tier datadissemination in large-scale wireless sensor networks, Wireless Netw.,vol. 11, nos. 12, pp. 161175, Jan. 2005.

[20] X. Chen and M. Xu, A geographical cellular-like architecture forwireless sensor networks, in Proc. Mobile Ad-Hoc Sensor Netw., 2003,pp. 249258.

[21] L. Buttyn and P. Schaffer, Position-based aggregator node electionin wireless sensor networks, Int. J. Distrib. Sensor Netw., vol. 2010,pp. 115, 2010.

[22] L. Buttyn and P. Schaffer, Position-based aggregator node electionin wireless sensor networks, Int. J. Distrib. Sensor Netw., vol. 2010,pp. 115, Jan. 2010.

[23] M. Eslaminejad and S. A. Razak, Fundamental lifetime mechanismsin routing protocols for wireless sensor networks: A survey and openissues, Sensors, vol. 12, no. 10, pp. 1350813544, Jan. 2012.

[24] W. B. Heinzelman, A. P. Chandrakasan, S. Member, andH. Balakrishnan, An application-specific protocol architecture forwireless microsensor networks, IEEE Trans. Wireless Commun., vol. 1,no. 4, pp. 660670, Oct. 2002.

[25] A. Manjeshwar and D. P. Agrawal, TEEN: A routing protocolfor enhanced efficiency in wireless sensor networks, in Proc. 15thInt. Parallel Distrib. Process. Symp. (IPDPS), vol. 1. Apr. 2000,pp. 20092015.

[26] A. Manjeshwar and D. P. Agrawal, APTEEN: A hybrid protocol forefficient routing and comprehensive information retrieval in wireless,in Proc. 16th Int. Parallel Distrib. Process. Symp., 2002, p. 48.

[27] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, Energy-efficient communication protocol for wireless microsensor networks,in Proc. 33rd Annu. Hawaii Int. Conf. Syst. Sci., Jan. 2000.

Abdul Waheed Khan received the Degree fromthe Department of Computer Science, Universityof Peshawar, Peshawar, Pakistan, in 2005, and theM.Sc. (Hons.) degree in digital communicationsnetworks from London Metropolitan University,London, U.K., in 2008. He is currently pursuing thePh.D. degree at the Faculty of Computing, Univer-sity of Technology Malaysia, Johor Bahru, Malaysia.From 2009 to 2012, he was a Lecturer with theFaculty of Computing and Information Technologyin Rabigh, King Abdulaziz University, Jeddah, Saudi

Arabia. His research areas include wireless sensor networks, ad hoc networks,and next-generation networks.

Abdul Hanan Abdullah received the Ph.D. degreefrom Aston University, Birmingham, U.K., in 1995.He is currently a Professor with the Univer-sity of Technology Malaysia (UTM), Johor Bahru,Malaysia. From 2004 to 2011, he was the Dean ofthe Faculty of Computer Science and InformationSystems. He is currently heading the Pervasive Com-puting Research Group under K-Economy ResearchAlliances, UTM. His research interests include wire-less sensor networks, mobile ad hoc networks, net-work security, and next-generation networks.

Mohammad Abdur Razzaque (M12) received theB.Sc. degree in applied physics, the M.Sc. degree inelectronics and communication engineering, and thePh.D. degree in computer science. He is currentlya Senior Lecturer with the Faculty of Computing,University of Technology Malaysia, Johor Bahru,Malaysia. His research interests include wirelessand mobile computing and communications, cloudcomputing, and big data analytics. He has authoredseveral peer-reviewed publications and has served asa regular reviewer for various international reputed

journals and conferences. He is a member of the IEEE Vehicular TechnologySociety, the IEEE Computer Society, and the Association for ComputingMachinery.

Javed Iqbal Bangash received the B.E. degreein computer information systems engineering fromthe University of Engineering and Technology,Peshawar, Pakistan, in 2005, and the M.Sc. (Hons.)degree in digital communications networks fromLondon Metropolitan University, London, U.K., in2008. He is currently pursuing the Ph.D. degree atthe Faculty of Computing, University of TechnologyMalaysia, Johor Bahru, Malaysia. He was a Lecturerat the College of Engineering, Majmaah University,Al Majmaah, Saudi Arabia. His research areas

include wireless body sensor networks, wireless sensor networks, and ad hocnetworks.

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 150 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 600 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages false /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 400 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/Description >>> setdistillerparams> setpagedevice

2014 VGDRA

Documents

Transcript of 2014 VGDRA